QAN App Get Filters API end point Response time could be improved
General
Escalation
General
Escalation
Description
How to test
None
How to document
None
Attachments
1
Smart Checklist
Activity
Show:

Puneet Kala February 8, 2022 at 9:27 AM
All nodes run on separate nodes, so this consumption of resources is just for pmm-server. I have to investigate with bigger boxes.

Roma Novikov February 8, 2022 at 9:22 AM
But, what is the resource usage for the Server without any request to QAN?
It might be - we consumed a lot of this 2CPU already by just running all DB's in the pmm server.
How the response time behaves on bigger boxes?
Details
Details
Assignee
Unassigned
UnassignedReporter

Priority
Components
Needs QA
Yes
Needs Doc
Yes
Smart Checklist
Open Smart Checklist
Smart Checklist

Open Smart Checklist
Created February 8, 2022 at 5:39 AM
Updated November 13, 2024 at 8:34 AM
User Impact: Bad QAN performance when applying filters or interacting with QAN app gives a feeling to not use this tool anymore.
PMM-QA set up another long-running instance with a good amount of queries executing on all 3 major databases.
2 Core, 4 GB RAM PMM-Server Host
The idea behind this was to observe the load timing for several components in the QAN app.
1) QAN overview, this seems to be loading in fair time when we have about 25k distinct queries it takes roughly 13 seconds to load the data, of course, there is the scope of improvement here, but we need to set this based on some parameters and metrics defined by PM and Engineering. /v0/qan/GetReport
2) What's not really obvious and acceptable is the amount of time Filters take to load, the moment you keep increasing the time range from last 12 hours default to last 7 days and then finally to last 30 days.
The load timing of QAN filters section, /v0/qan/Filters/Get keeps increasing for the same load with which the overview
Fetching response for last 7 days takes around 33 seconds, fetching last 30 days takes around 40 seconds to load just the filter section.